You can keep the list of the N biggest (with their sizes).
Something like this (this might crash as I can't test it but the idea is there:
from Bio import SeqIO
file="sequence.gb"
output=open("Top3.faa", "w")
record=SeqIO.parse(file, "genbank")
rec=next(record)
print('The genes with the top 3 longest lengths have beens saved in Top3.faa')
# Largest genes and their size, sorted from the shortest to the longest.
# size first, gene name next, then seq.
largest_genes = [ (0, None, None) ] * 3; # initialize with the number of genes you need.
for f in rec.features:
end = f.location.end.position
start = f.location.start.position
length = end-start
bug = (rec.seq)
if f.type=='CDS' and 'gene' in f.qualifiers:
if length > largest_genes[0][0]: # [0] gives the first, [0] gives the length.
# This one is larger than the smallest one we have.
largest_genes = largest_genes[1:] # drop the smallest one.
# add this one
largest_genes.append((length, f.qualifiers['gene'], str(bug[start:end])))
largest_genes.sort() # re-sort.
for length, name, seq in largest_genes:
# name is not used but available.
output.write('>')
output.write(str(lenth))
output.write('\n')
output.write(seq)
output.write('\n')
output.close()